Abstract
Suicide remains a major public health concern. Suicide and suicidal behaviours are statistically infrequent, even in populations at risk [1–3]. The central problem confronting clinicians working in suicide prevention is that while the total number of suicides in a population is pretty large, the risk of suicide to any given individual, even those with multiple risk factors, is pretty small.
Medical professionals deal with almost no other event as rare as suicide. Although suicidal ideation and attempts are associated with increased suicide risk, most individuals with suicidal thoughts or attempts will not die by suicide. Suicide attempts and ideation occur in approximately 0.7% and 5.6%, respectively, of the general US population [4]. In comparison, in the USA the annual incidence of suicide in the general population is approximately 11 suicides for every 100,000 persons, or 0.011% of the total population per year [5]. Even under the best conditions, the attempted prediction of an uncommon behaviour such as suicide unavoidably generates a huge number of false-positive and false-negative cases [6]. It has been shown that for realistic assumptions of sensitivity and specificity, a screening test for suicide risk would have a positive predictive value of 0.3% and generate a vast number of false positives [6]. Besides, the prediction of suicidal behaviour is based on inexact criteria that are relatively poor at predicting the behaviour of a given person, and suicidal individuals often conceal or deny suicidal thoughts in order to avoid unwanted intervention efforts, such as involuntary hospitalization.
The general public expects medical professionals to predict suicide attempts and to protect patients from death by suicide. However, the belief that clinicians can predict suicidal behaviour is unrealistic. For example, stepwise multiple logistic regression was utilized in an attempt to develop a statistical model that would predict suicide in a group of about 2000 inpatients with affective disorders [7]. The risk factors identified by this approach included the number of prior suicide attempts, suicidal ideation on admission, bipolar affective disorder (manic or mixed type), gender, outcome at discharge, and unipolar depressive disorder in individuals with a family history of mania. However, the model failed to identify any of the patients who committed suicide, which supported the contention that it is not possible to predict suicide, even among a high-risk group of inpatients.
Although predicting suicide at the level of individual patients is not possible at the present time, clinicians are obligated to make everything possible to prevent suicidal behaviour in their patients. It is important to note that our concern about prediction of suicide is related to our failure to prevent all suicides. If suicide is hard to predict, its prevention is even harder to identify. It is likely that many suicidal individuals are recognized and successfully treated.
At the present time, our knowledge of suicide prevention and prediction is very limited. Innovative psychobiological studies that can help to identify better criteria for prediction of suicide are critically needed.